(19)
(11)EP 3 309 521 B1

(12)EUROPEAN PATENT SPECIFICATION

(45)Mention of the grant of the patent:
29.07.2020 Bulletin 2020/31

(21)Application number: 16193922.8

(22)Date of filing:  14.10.2016
(51)International Patent Classification (IPC): 
G01F 1/32(2006.01)

(54)

METHOD FOR EVALUATING A FREQUENCY SPECTRUM

VERFAHREN ZUR BEWERTUNG EINES FREQUENZSPEKTRUMS

PROCÉDÉ D'ÉVALUATION D'UN SPECTRE DE FRÉQUENCES


(84)Designated Contracting States:
AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

(43)Date of publication of application:
18.04.2018 Bulletin 2018/16

(60)Divisional application:
20180163.6

(73)Proprietor: Grundfos Holding A/S
8850 Bjerringbro (DK)

(72)Inventor:
  • Pedersen, Ole Hartlev
    8850 Bjerringbro (DK)

(74)Representative: Patentanwälte Vollmann Hemmer Lindfeld Partnerschaft mbB 
Wallstraße 33a
23560 Lübeck
23560 Lübeck (DE)


(56)References cited: : 
EP-A2- 2 388 614
WO-A2-2013/067338
US-B2- 8 215 183
WO-A1-90/04230
JP-B2- 3 675 443
  
      
    Note: Within nine months from the publication of the mention of the grant of the European patent, any person may give notice to the European Patent Office of opposition to the European patent granted. Notice of opposition shall be filed in a written reasoned statement. It shall not be deemed to have been filed until the opposition fee has been paid. (Art. 99(1) European Patent Convention).


    Description


    [0001] The invention relates to a method for evaluating a frequency spectrum, a method for determining flow of a vortex measuring device, and a vortex sensor for carrying out the method for evaluating a frequency spectrum.

    [0002] Measuring devices, as for example, vortex flow meters are known in prior art for measuring flow velocity of liquids or gases in pipelines. Further, there are known so-called multivariable vortex meters which offer a couple of advantages over regular vortex flow meters. Specifically, these multivariable vortex meters are temperature and pressure compensated, thereby offering more precise results and enabling direct measurement of mass flow in fluids flowing through a pipeline.

    [0003] Basically, a multivariable vortex meter may be configured to measure flow, relative pressure and temperature, for example, by means of a 3-in-one sensor with a microelectromechanical system chip (MEMS). This sensor chip has two measuring components:
    1. 1. A Wheatstone bridge of four resistors with piezo-resistive effect at a pressure sensitive membrane. If a differential pressure is applied, the membrane will bend, and the four resistors in the Wheatstone bridge are subject to mechanical stress and will change resistance with two resistors decreasing resistance and two resistors increasing resistance. Under application of a DC voltage, the Wheatstone bridge will yield a pressure dependent voltage output.
    2. 2. A resistor with a Temperature Coefficient of Resistance (TCR) changing due to the temperature of the MEMS.


    [0004] In order to measure flow, relative pressure and temperature of a fluid flowing through a pipeline by means of only one MEMS element, the sensor may only have one open port at the positive MEMS side, as used in relative pressure sensors. At the negative pressure side, atmospheric pressure is applied from the inside of the sensor housing.

    [0005] However, for measuring flow by means of vortex meters, usually a two-port sensor is implemented being arranged downstream of the bluff body. The velocity of the vortices causes small pressure changes, and when passing the two-port sensor, the latter will measure a small differential pressure. While vortices will pass the sensor in an alternating manner with respect to the positive and negative pressure ports, the signal measured by the sensor will be a small sine signal comprising some noise from hydraulic or other noise sources. The number of the vortices generated per second is proportional to the flow so that the flow algorithm has to determine the frequency of the sine and multiply it with a factor to obtain a value for the flow in the desired unit.

    [0006] The use of two-port sensors in vortex flow meters has the advantage that a rather good sine signal comprising little noise is obtained from the sensor. With low flow, the sine amplitude will have very low amplitude, but the relative pressure (namely, pressure related to atmospheric pressure) in the vortex tube will comprise hydraulic noise as harmonic noise components and/or large and rapid pressure deviations. Thus, there is a disadvantage with respect to the configuration described above that the two-port pressure sensor is only sensitive to differential pressure but not to relative pressure, because the two-port solution effectively protects the sine signal from being altered by the hydraulic noise of the relative pressure.

    [0007] Nevertheless, the above described two-port solution is able to measure both flow and temperature. In order to be able to additionally measure relative pressure by means of the above mentioned MEMS, the two-port sensor housing has to be replaced by a one-port sensor housing.

    [0008] However, when using a one-port housing, the sensor will also be sensitive to hydraulic noise emanating from the hydraulic relative pressure noise. Thus, the resulting sine curve will be modified or distorted due to that noise. In fact, since the characteristics and magnitude of the hydraulic noise differ from application to application, and even will differ over time within the same application, this unpredictable noise poses a severe problem to measuring vortex flow with only one pressure port.

    [0009] Also other measuring devices, as for example, thermal flow meters comprising a plurality of temperature sensors for measuring a thermal profile around a heated body have to deal with the problem of unpredictable noise. With respect to such applications, it is necessary to measure very small temperature differences. Noise, for example, from an electronic circuit, can be problematic as well as noise caused by the medium.

    [0010] EP 2 388 614 A2 discloses a laser Doppler velocimeter using self-mixing amplification of back-reflections from scatterers below the surface of a flow. A signal-to-noise ratio is enhanced through signal processing techniques.

    [0011] JP 3675443 B2 describes a Fast-Fourier transformation processing of a detection signal for identifying peaks in a peak envelope curve after removal of a noise component.

    [0012] Therefore, the present invention is based on the object to provide a method for measuring, managing, and reducing noise during measurements which are disturbed by relatively high noise. This object is solved according to the present invention by a method for evaluating a frequency spectrum representative of at least one time-dependent signal having the features according to claim 1, a method for determining flow of a vortex measuring device having the features according to claim 7, and a vortex sensor for carrying out the method having the features according to claim 13. Preferred embodiments of the present invention are specified in the respective dependent claims.

    [0013] Accordingly, a method for evaluating a frequency spectrum representative of at least one time-dependent signal is provided, the at least one time dependent signal being derived from an output from a measuring device under predetermined measuring device operating conditions, the time dependent signal comprising a portion being representative of a wanted signal, and a portion being representative of noise, wherein the method comprises the steps of
    • determining, based on the frequency spectrum of the signal a representative of the noise floor,
    • identifying, based on the frequency spectrum of the signal derived under the predetermined operating condition, a peak component, and
    • if the peak component satisfies a relative peak criterion determined on the basis of the determined representative of the noise floor, determining the wanted signal by applying a predetermined algorithm.


    [0014] By the inventive method, measurements which are disturbed by relatively high noise and very small local differences smaller than the noise can be carried out reliably. The inventive method enables measurements adaptive to the current noise which makes an auto noise reduction by an averaging procedure adapted to the current situation which only delivers a measurable difference, if there is a statistically significant difference. By the inventive method, it is possible to measure very small differences that normally cannot be measured when the signal is "buried" in or mixed by noise. The inventive method applies noise reduction that is only activated when there is a need for it, thus, the noise reduction is adaptive and fits the specific noise. In applications where the noise level changes from time to time, the inventive method can be used to secure that no ensuring is biased due to noise. When the noise is low, the resolution is high, and when the noise is high, the resolution is low. The inventive method ensures that the measurement is set to zero whenever there is no statistical significance that it does not have a value greater than zero.

    [0015] Further, according to the present invention, a method is provided, wherein the measuring device is a vortex measuring device, and wherein the time-dependent signal comprises a portion being representative of a flow, and a portion being representative of noise. Thus, a reliable procedure for distinguishing between noise component and a vortex component in a spectrum, in particular, generated at low flow rates is provided rendering correct results for any application of a vortex sensor implemented for flow measurement in a pipe. In particular, the inventive method is very well suited to be used in connection with vortex meters comprising a one-port housing so as to be able to determine, even at low flow rates, relative pressure, temperature and flow with only one MEMS.

    [0016] Alternatively, the measuring device may be a thermal flow meter comprising a plurality of temperature sensors for measuring a thermal profile around a heated body. Also in this kind of application, it is necessary to measure the very small temperature differences and this embodiment enables measuring very low flow. In contrast to most flowmeter principles, the method according to this preferred embodiment is useful by employing an adaptive noise reduction.

    [0017] According to the invention, a value for the noise floor is calculated as an average of a number of selected noise representative frequency spectrum components. Thereby, the accuracy of the results is increased substantially.

    [0018] According to the invention, the noise representative frequencies spectrum components are selected among frequency components below a predefined frequency value, wherein the noise representative frequencies spectrum components are the remaining components after selecting a predetermined number of high value frequency components, wherein the high value frequency components are a predetermined number of frequency components with the highest amplitude values.

    [0019] Preferably, a relative peak signal is determined from the difference between the noise floor and the wanted signal, wherein the relative peak signal is larger than or equal to the relative peak criterion.

    [0020] It is advantageous, if the predetermined algorithm is adaptive with respect to the noise by
    • always calculating the relative peak criterion as a function of the noise being present, and
    • selecting a suitable averaging procedure securing a stable wanted signal output.


    [0021] Thereby, an intelligent adaptive solution is provided being suitable and providing reliable results for any kind of application.

    [0022] The frequency spectrum may be calculated by averaging frequency spectra of a predetermined number of time dependent signals derived from the output of the measuring device. This increases the accuracy of the results even further.

    [0023] Further, a method for determining flow of a vortex measuring device is provided, comprising the step of computing a sequence of frequency spectra, each frequency spectrum of the sequence being computed by the method described above, wherein for the first spectrum of the sequence, the predetermined number is 1, and for each subsequent spectrum of the sequence, the predetermined number is increased by a predetermined positive integer value, and wherein for each frequency spectrum of the sequence of frequency spectra, the corresponding flow is determined by applying the predetermined algorithm, and evaluating if at least one flow is determined to be greater than the noise floor, then the first spectrum of the sequence of spectra where the determined flow is greater than the noise floor, is selected as basis for calculating the flow of the vortex measuring device, otherwise determining the flow to be zero flow.

    [0024] Preferably, each frequency spectrum is divided into a number of discrete frequency components, and identifying the peak component comprises the step of deriving the peak by interpolating based on several frequency components.

    [0025] According to a preferred embodiment, each frequency spectrum of the sequence of frequency spectra is numbered with a consecutively increasing channel number, the first frequency spectrum being numbered with the smallest channel number, the last frequency spectrum being numbered with the largest channel number, wherein frequency spectra having higher channel numbers being averaged over frequency spectra of predetermined numbers of time dependent signals with lower frequency limits than frequency spectra of predetermined numbers of time dependent signals, having lower channel numbers, wherein prior to the evaluating step, for each frequency spectrum of the sequence of frequency spectra, the corresponding flow is determined by applying the predetermined algorithm, and if the flow of the spectrum having the smallest channel number is determined to be greater than the noise floor, that spectrum is selected as basis for calculating a steady state flow of the vortex measuring device, otherwise if the flow of the spectrum having the smallest channel number is determined to be equal to the noise floor and at least one of the spectra having a channel number between the smallest and the largest channel number having a flow greater than the noise floor, one of the spectra having a flow greater than the noise floor is selected as a basis for calculating a dynamic flow of the vortex measuring device, otherwise performing the evaluating step.

    [0026] The method may further comprise a step of determining a stationary flow and a dynamic flow.

    [0027] Preferably, the output signal is a relative pressure signal.

    [0028] Moreover, it is advantageous, if the vortex measuring device comprises a vortex sensor which is arranged within a one-port housing.

    [0029] Further, according to the invention, a vortex sensor is provided being adapted to carry out the method according to any one of the preceding claims.

    Brief description of the figures



    [0030] Further details and features of the invention as well as concrete embodiments of the invention may be derived from the following description in connection with the drawing, in which:
    Fig. 1A, Fig. 1B
    respectively show a spectrum comprising a vortex component at 28 Hz with Fig. 1A showing a frequency spectrum being obtained after one run and Fig. B showing a frequency spectrum being obtained after averaging over sixteen runs;
    Fig. 2A, Fig. 2B
    respectively show the development of respective components from the frequency spectrum develop after different numbers of averaging procedures, wherein Fig. 2A refers to a vortex frequency component and Fig. 2B refers to a noise frequency component;
    Fig. 3
    shows a table for the basic selection of average channels;
    Fig. 4
    shows a high resolution frequency spectrum which illustrates an example for limiting the frequency area of average channels; and
    Fig. 5
    shows a table according to which steady state flow and dynamic flow are distinguished from a pattern of two valid peaks by using all average channels.


    [0031] Fig. 1A and Fig. 1B illustrate how the noise reduction system according to the present invention works on the basis of a frequency spectrum comprising a vortex component at 28 Hz.

    [0032] Prior to explaining Fig. 1A and Fig. 1B, however, the following basic explanations concerning measurements of flow, relative pressure, and temperature are given below.

    [0033] The temperature measurement has its own TCR (Temperature Coefficient of Resistance) component with respect to the MEMS. The flow measurement and the measurement of the relative pressure share the same Wheatstone bridge located at the pressure sensitive membrane of the MEMS. The Wheatstone bridge outputs the differential pressure applied across the MEMS. This pressure consists of an alternating signal (AC) to be used for the flow sensor application, and a steady state signal (DC) for the measurement of pressure, or a slow changing DC signal for the measurement of relative pressure.

    [0034] For the flow algorithm the AC signal is used but the DC signal, however, is not suited. Thus, a digital high pass filter provides a separate signal channel for the flow algorithm with only the AC signal and the DC signal being removed.

    [0035] The pressure algorithm uses the DC signal, but here, the AC signal is not suited. Thus, a digital low pass filter provides a separate signal channel for the pressure algorithm with only the DC signal and the AC signal being removed.

    [0036] Then, the relative pressure measured has to be compensated for dynamic pressure, because the diameter of the pipe through which the flow to be measured occurs has another diameter at location at which the MEMS is arranged.

    [0037] Further, the basis measurement with respect to flow algorithm is a measurement of the frequency of the vortex generated sine. The method, which is chosen in the embodiment described here, is based on a calculation of the frequency spectrum of the AC signal. This is a discrete spectrum comprising, e.g., 100 components spaced apart from each other by 4 Hz. On this basis, since each frequency will "see" 4 Hz to both sides, any frequency of a sine will be detected by one or two frequency components. If the vortex consists, for example, of one pure sine of e.g., 80 Hz and having an amplitude of 100 (e.g., 100 AD count of e.g., 1 mV, an amplitude of 100 mV), then the frequency components of 80 Hz in the spectrum will output the amplitude of 100, and all other components will provide an output of zero. If, e.g., noise of 30 Hz and an amplitude of 12 is also present, then the frequency component for 12 Hz will provide the output of the value 12 as well.

    [0038] Usually, a typical frequency spectrum will output a value above zero for the noise floor for all frequency components, even when the flow is zero. This effect results from electrical and hydraulic noise. When flow is applied to the flow meter, then a frequency component for the vortex sine will be output having a value above the value the noise floor has. Further, also some noise components may be present with a value higher than the value of the noise floor. Thus, the task of the flow algorithm in the sense of software is to analyze the frequency spectrum in order to be able to distinguish between the vortex sine component and the noise components. Since the vortex sine frequency flow characteristic is known, the latter is able to effectively detect and reject most of the noise components.

    [0039] The nature of noise within the frequency spectrum can be described as outlined below. If there is no flow, then the output of all frequency components will be a random value higher or less than a mean value. The means value represents the noise floor and the random behavior is almost a normally distributed function. Then, the random behavior can be described by a single parameter, namely, the standard spread or just the sigma. 68 % of the random output lies within one sigma and 99,7 % lies within 3 sigma.

    [0040] The noise amount and characteristic varies from application to application at the customer side, and so do noise floor and sigma. However, the sigma is represented by a certain known function of the noise floor so that when measuring the noise floor, then sigma may be calculated. Since the noise floor is measured, then, the actual noise floor and the actual sigma of the noise are also known for each application.

    [0041] If the application involves high hydraulic noise, then this noise will "eat" the lower frequency components. In this case, the minimum flow cannot be measured and the system avoids detecting noise as flow.

    [0042] Now, the basics for finding and approving the peak for the vortex will be described. First, the flow is analyzed with respect to the frequency spectrum so as to find the highest peak respectively meeting absolute peak criteria and relative peak criteria. The absolute peak criteria serve for verifying that the found peak is not a random noise signal but rather matches the characteristics of a vortex signal. The relative peak, which verifies that the found signal is not a random sign is determined by determining the difference between the maximum peak which has been found and the noise floor, whereby it has to have a minimum value, which is adopted to the actual sigma. The noise floor is calculated as the average of the peaks surrounding spectrum components. The frequency component being tested and its two neighbors are not included in the noise floor measurement, since the noise floor then would be higher than the correct value for a vortex component being tested.

    [0043] The following description is given for explaining how to distinguish between the flow and noise. The basic principle for distinguishing between components comprising noise and the components needed for the evaluation ("good" components) starts at a condition of zero flow at which only noise being present, and at which the sigma is highest. The frequency spectrum component responds to noise with the noise floor (mean value) and the dispersion, sigma, which is calculated as described above. The flow algorithm, therefore, will always include both, noise floor and sigma. First, the flow algorithm is based on a zero flow hypothesis test. If the component output lies within the noise sigma (e.g., 3 sigma), this means that the relative peak criteria is not met, but the zero flow hypothesis is met with the flow being zero. Otherwise, if the absolute and relative peak criteria are met, then the frequency component output refers to flow.

    [0044] The basic concept of the zero flow hypothesis in summary is as follows: The noise floor is measured, the sigma of the noise is calculated, and the relative peak is calculated by input of a confidence level and sigma. The zero flow hypothesis is selected with a confidence level (e.g. 3 sigma), thereby making it unlikely that noise is able to trigger an output above the relative peak criteria when flow is zero. If an output exceeds the relative peak criteria, it is likely that flow is present. In this case, flow will be detected, if the absolute peak criteria also are met.

    [0045] The appropriate level of the relative peak criteria is the sigma being multiplied by a factor providing the desired confidence (e.g. a factor 3 will provide 3 sigma as confidence level). This is configurable. However, if the confidence level is set to be higher, then the relative peak criteria will increase. Thus, if sigma increases, the relative peak criteria increase too. The relative peak criteria respond to the actual noise in the application so that the hypothesis test parameters actually adopt to the actual noise in the application. This provides for a reliable and effective procedure to distinguish between "too noise components" and "good components" in an unpredictable application.

    [0046] Now, the basics of the noise reduction system will be described. The noise is reduced by averaging the frequency components over several runs N. When averaging, the sigma σ of the average result is reduced to the value

    When averaging, for example, over a series of four measurements, the sigma of the result is reduced to half, whereas averaging over a series of sixteen measurements reduces sigma to ¼. Both components and the noise floor are averaged.

    [0047] Fig. 1A shows a spectrum for only one run and vortex component at 28 Hz, whereas Fig. 1B shows a spectrum for an average of sixteen runs and a vortex component at 28 Hz. After separating the 28 Hz vortex component from the noise, the relative peak value, which in Fig. 1B is also reduced to ¼ compared to Fig. 1A, now allows the 28 Hz vortex component to always be selected. By repeating the calculation of the frequency components and averaging the latter, the sigma of the dispersion is reduced. This, however, only is necessary in the low flow range, as already explained above. Channels of an average of 2, 4, 8, and 16 are available, whereby the last channel has ¼ sigma of the channel without averaging.

    [0048] The relative peak criteria are calculated for each channel including the sigma and noise floor measurement as well as the sigma of each channel. The sigma of the noise floor is rather low since it is measured by six different components reducing the sigma to 0.4 * sigma. Each channel has its own noise floor due to different averaging.

    [0049] With respect to Fig. 1A, the frequency spectrum is shown without an averaging procedure for a vortex of 28 Hz. The upper light grey colored portion of each column represents the respective portion of noise, which in the example of Fig. 1A is rather large. Thus, regarding the highest peak at a frequency of 28 Hz, it cannot be determined with certainty whether this corresponds to the vortex. The relative peak criteria is only suitable for distinguishing between the 28 Hz vortex component and the surrounding noise components so as to avoid a misinterpretation of a noise peak for the vortex peak. However, here, the 28 Hz peak has a low probability to be selected, thus, the flow algorithm, without the average channel system and without the adaptive relative peak criteria, will indicate zero flow most of the time and will only be able to determine the correct 28 Hz vortex component for about 1/5 of times according to standard probability.

    [0050] In contrast, in Fig. 1B the frequency components are averaged sixteen times, thereby improving the detection of the vortex component in the spectrum. In this example, the vortex component will be determined correctly from the spectrum with a 100 % probability.

    [0051] The selection of the average channel is tested from the "no-average channel", followed by the "average-of-2-channel", and so on up to the last "average-of-16-channel". Then, the first channel detecting flow is selected. If no channel at all detects flow, then the flow is determined to be zero.

    [0052] The above described procedure is very reliable due to the averaging, and thus, there is no risk to select a noise component wrongly instead of the desired vortex component from the spectrum.

    [0053] The noise reduction system, thereby, is adaptive both to noise of each application and to the appropriate amount of averaging.

    [0054] Fig. 2A shows how a vortex frequency component develops according to the number of averaging procedures, and Fig. 2B shows how a noise frequency component develops over the number of averaging procedures.

    [0055] It should be noted that the sine signal amplitude generated by the vortices increase by a magnitude of two with flow. As noise usually will be rather constant in magnitude, the ratio of signal to noise (S/N) is very low, thus resulting in a noise signal with low flow. However, it will quickly increase with flow, and the sine curve will be free from critical noise rather fast. Therefore, the problem with respect to noise focusses to the lowest area of the flow range.

    [0056] In Fig. 2A, as an example, the 28 Hz component shown in Fig. 1B is used, whereby the 28 Hz component, the relative peak criteria, and the noise floor spread or dispersion are shown as a function of the number of averaging procedures, starting from 1 to 16 runs. Here, the average of 16 is the only one which enables a correct determination of the vortex component. It should be noted that the knowledge that a vortex component will always have lower noise dispersion than a noise component contributes to the reliability of the hypothesis test.

    [0057] Fig. 2B shows the spectrum at zero flow. Also here, the noise frequency component, the relative peak criteria and the noise floor spread or dispersion are shown as a function of the number of averaging procedures. The range of noise with the desired confidence level is below the relative peak criteria at all averaging levels so as to provide the same confidence for zero flow output for every averaging level.

    [0058] Fig. 3 shows a table for the basic selection of average channels. In the table, a value of "0" means no valid peak in the average channel. In contrast, in the table, a value of "1" indicates a valid peak in the average channel. A value of "X" means, that the latter should not be considered. A valid peak can be defined as output which meets the absolute peak criteria and the relative peak criteria as well. The selected average channel here is indicated by the grey color. This basic selection prioritizes the lowest average channels. However, a more advantageous selection will be described below.

    [0059] The vortex frequency is calculated as an interpolated value of several frequency components within the spectrum. This provides an accurate result at any vortex frequency. This interpolating accuracy will even increase with increasing averaging runs, when flow is in steady state, since the components comprise less noise and will have lower tolerance. Thus, preferably, the channels with high averaging should be used, if possible.

    [0060] The minimum noise sets a minimum limit for the frequency for each average channel. This minimum noise will not be able to provide the lowest frequency component in the spectrum being valid for each measurement. Thus, this channel is not allowed to measure the lowest flow area. The "average-of-two-channel" is able to reliably determine lower frequencies than the "no-average-channel" and therefore represents the lower limit. The same applies for all other average channels. The "average-of-sixteen-channel" starts at the lowest frequency.

    [0061] Fig. 4 shows a high resolution frequency spectrum which illustrates an example for limiting the frequency area of average channels. Here, the lowest frequency in the spectrum is component 3 at 20 Hz, whereas components 0, 1, and 2 at 8 Hz, 12 Hz, and 16 Hz, respectively, are below the frequency range of the vortex frequency range. The individual limit of the frequency range for the average channels provides for better interpolation of frequency accuracy and better suppression of sporadic noise.

    [0062] Even better results and accuracy as well as an even better suppression of sporadic noise can be achieved by limiting the step from one average channel to the next one, and especially the step over more than one channel. This is of particular relevance with respect to steady state flow, and in those cases, where an "average-of-16-channel" has to be used. In this case, sporadic valid peaks of the low average channels can be avoided to be selected. This procedure is implemented in the so-called intelligent average channel selection described below and illustrated in Fig. 5.

    [0063] Fig. 5 shows a table according to which "steady state flow" and "dynamic flow" may be distinguished from a pattern of two valid peaks by using all average channels, thereby providing an intelligent average channel selection procedure according to the respective state. Namely, at steady state flow, the "average-of-16-channel" will show valid peaks, but no dynamic flow. The table of Fig. 5 shows the patterns to be detected. A table value of "0" indicates that no valid peak is present for any component in the respective average channels. A table value of "1" indicates that a valid peak is present for at least one component in the respective average channels, whereas the value "X" indicates that this should not be taken into consideration. A valid peak is defined as output which meets the absolute peak criteria and the relative peak criteria. The cells of the table which are shaded in grey indicate the selected average channels.

    [0064] At steady state flow, the "average-of-16-channel" is selected to suppress sporadic noise and high accuracy. However, if at least three lower average channels above also have valid peaks, then the "average-of-8-channel" is selected to enable faster response.

    [0065] At dynamic flow, the "average-of-2-channel" is selected in order to suppress sporadic noise and still provide fast response time. If, however, a pattern which is not included in the table shown in Fig. 5 is present, then the selection of default average channels as illustrated in the table shown in Fig. 3 is to be applied. However, the above procedure described above with respect to Fig. 5 compared to the procedure described with respect to Fig. 3 has the advantage that by always selecting a larger averaging here in contrast to always select the smallest averaging (Fig. 3), an output having less dispersion/spread is obtained. Accordingly, the result is more resistant against large and sporadic noise pulses that may "cheat/trick" the lowest averaged channels.

    [0066] Thus, the above described procedure provides an intelligent noise reduction means integrated into the sensor algorithm for measuring, managing, and reducing noise when carrying out a measurement with a measuring device having more than one senor. As already mentioned above, the method according to the present invention is suitable for measurement devices as vortex measurement devices or thermal flow meters. However, also other measurement devices having a plurality of sensors are can be employed.

    List of reference numerals



    [0067] 
    1
    noise floor average
    2, 2'
    relative peak min criteria
    3, 3'
    absolute peak min criteria
    4, 4'
    absolute peak 3 σ spread
    5, 5'
    noise component 3 σ spread
    6, 6'
    noise floor 3 σ spread
    7
    noise spread, three sigma
    8
    relative peak limit for zero flow/flow
    9
    noise floor spread, three sigma
    10
    noise sensitive area
    11
    no average
    12
    average of 2 outputs
    13
    average of 4 outputs
    14
    average of 8 outputs
    15
    average of 16 outputs



    Claims

    1. Method for evaluating a frequency spectrum representative of at least one time-dependent signal, the at least one time dependent signal being derived from an output from a measuring device under predetermined measuring device operating conditions, the time-dependent signal comprising a portion being representative of a wanted signal, and a portion being representative of noise, wherein the method comprises the steps of

    - determining, based on the frequency spectrum of the signal, a value representative of the noise floor,

    - identifying, based on the frequency spectrum of the signal derived under the predetermined operating condition, a peak component, and

    - if the peak component satisfies a relative peak criterion determined on the basis of the determined value representative of the noise floor, determining the wanted signal by applying a predetermined algorithm,

    wherein the value representative of the noise floor is calculated as an average of a number of selected noise representative frequencies spectrum components,
    characterised in that
    the noise representative frequencies spectrum components are selected among frequency components below a predefined frequency value, wherein the noise representative frequencies spectrum components are the remaining components after selecting a predetermined number of high value frequency components, wherein the high value frequency components are a predetermined number of frequency components with the highest amplitude values.
     
    2. The method according to claim 1, wherein a relative peak signal is determined from the difference between the noise floor and the wanted signal, wherein the relative peak signal is larger than or equal to the relative peak criterion.
     
    3. The method according to any one of the preceding claims, wherein the predetermined algorithm is adaptive with respect to the noise by

    - always calculating the relative peak criterion as a function of the noise being present, and

    - selecting an averaging procedure with a number of averaging runs, wherein the number of averaging runs is based on the current noise floor.


     
    4. The method according to any one of claims 1 to 3, wherein the frequency spectrum is calculated by averaging frequency spectra of a predetermined number of time dependent signals derived from the output of the measuring device.
     
    5. The method according to any one of the preceding claims, wherein the measuring device is a vortex measuring device, and wherein the time-dependent signal comprises a portion being representative of a flow, and a portion being representative of noise.
     
    6. The method according to any one of claims 1 to 4, wherein the measuring device is a thermal flow meter comprising a plurality of temperature sensors for measuring a thermal profile around a heated body.
     
    7. A method for determining flow of a vortex measuring device, comprising the step of computing a sequence of frequency spectra, each frequency spectrum of the sequence being computed by the method of claim 4, wherein for the first spectrum of the sequence, the predetermined number is 1, and for each subsequent spectrum of the sequence, the predetermined number is increased by a predetermined positive integer value, and
    wherein for each frequency spectrum of the sequence of frequency spectra, the corresponding flow is determined by applying the predetermined algorithm, and
    evaluating if at least one flow is determined to be greater than the noise floor, then the first spectrum of the sequence of spectra where the determined flow is greater than the noise floor, is selected as basis for calculating the flow of the vortex measuring device,
    otherwise determining the flow to be zero flow.
     
    8. The method of claim 7, wherein each frequency spectrum is divided into a number of discrete frequency components, and identifying the peak component comprises the step of deriving the peak by interpolating based on several frequency components.
     
    9. The method of claim 7, wherein each frequency spectrum of the sequence of frequency spectra is numbered with a consecutively increasing channel number, the first frequency spectrum being numbered with the smallest channel number, the last frequency spectrum being numbered with the largest channel number, wherein frequency spectra having higher channel numbers being averaged over frequency spectra of predetermined numbers of time dependent signals with lower frequency limits than frequency spectra of predetermined numbers of time dependent signals, having lower channel numbers,
    wherein prior to the evaluating step of claim 7, for each frequency spectrum of the sequence of frequency spectra, the corresponding flow is determined by applying the predetermined algorithm, and

    if the flow of the spectrum having the smallest channel number is determined to be greater than the noise floor, that spectrum is selected as basis for calculating a steady state flow of the vortex measuring device,

    otherwise if the flow of the spectrum having the smallest channel number is determined to be equal to the noise floor and at least one of the spectra having a channel number between the smallest and the largest channel number having a flow greater than the noise floor, one of the spectra having a flow greater than the noise floor is selected as a basis for calculating a dynamic flow of the vortex measuring device,

    otherwise performing the evaluating step of claim 7.


     
    10. The method according to any one of the preceding claims, further comprising a step of determining a stationary flow and a dynamic flow.
     
    11. The method of any one of the claims 1 to 5 or 7 to 10, wherein the output of the measuring device represents a relative pressure between the at least two sensors.
     
    12. The method of any one of the claims 5 or, 7 to 9, wherein the vortex measuring device comprises a vortex sensor which is arranged within a one-port housing.
     
    13. A vortex sensor adapted to carry out the method according to any one of the claims 1 to 5 or 7 to 12.
     


    Ansprüche

    1. Verfahren zur Evaluierung eines Frequenzspektrums, das für mindestens ein zeitabhängiges Signal repräsentativ ist, wobei das mindestens eine zeitabhängige Signal von einem Ausgang aus einer Messvorrichtung unter vorbestimmten Betriebsbedingungen der Messvorrichtung abgeleitet wird, wobei das zeitabhängige Signal einen Abschnitt, der für ein gewünschtes Signal repräsentativ ist, und einen Abschnitt, der für ein Rauschen repräsentativ ist, umfasst, wobei das Verfahren die folgenden Schritte umfasst:

    - Bestimmen, auf der Basis des Frequenzspektrums des Signals, eines Werts, der für das Grundrauschen repräsentativ ist,

    - Identifizieren, auf der Basis des Frequenzspektrums des Signals, das unter der vorbestimmten Betriebsbedingungen abgeleitet wird, einer Spitzenkomponente, und

    - wenn die Spitzenkomponente ein relatives Spitzenwertkriterium erfüllt, das auf der Basis des vorbestimmten Werts bestimmt wird, der für das Grundrauschen repräsentativ ist, Bestimmen des gewünschten Signals durch Anwenden eines vorbestimmten Algorithmus,

    wobei der Wert, der für das Grundrauschen repräsentativ ist, als Mittel einer Anzahl ausgewählter, für ein Rauschen repräsentativer Frequenzspektrumkomponenten berechnet wird,
    dadurch gekennzeichnet, dass die für ein Rauschen repräsentativen Frequenzspektrumkomponenten aus Frequenzkomponenten unter einem vordefinierten Frequenzwert ausgewählt werden, wobei die für ein Rauschen repräsentativen Frequenzspektrumkomponenten die verbleibenden Komponenten nach der Auswahl einer vorbestimmten Anzahl von Frequenzkomponenten mit einem hohen Wert sind, wobei die Frequenzkomponenten mit einem hohen Wert eine vorbestimmte Anzahl von Frequenzkomponenten mit den höchsten Amplitudenwerten sind.
     
    2. Verfahren nach Anspruch 1, wobei ein relatives Spitzensignal aus der Differenz zwischen dem Grundrauschen und dem gewünschten Signal bestimmt wird, wobei das relative Spitzensignal größer als oder gleich dem relativen Spitzenwertkriterium ist.
     
    3. Verfahren nach einem der vorhergehenden Ansprüche, wobei der vorbestimmte Algorithmus in Bezug auf das Rauschen adaptiv ist durch

    - immer Berechnen des relativen Spitzenwertkriteriums als Funktion des Rauschens, das vorhanden ist, und

    - Auswählen einer Mittelwertbildungsprozedur mit einer Anzahl von Mittelwertbildungsdurchläufen, wobei die Anzahl von Mittelwertbildungsdurchläufen auf dem aktuellen Grundrauschen basiert.


     
    4. Verfahren nach einem der Ansprüche 1 bis 3, wobei das Frequenzspektrum durch eine Mittelwertbildung von Frequenzspektren einer vorbestimmten Anzahl von zeitabhängigen Signalen berechnet wird, die von dem Ausgang der Messvorrichtung abgeleitet werden.
     
    5. Verfahren nach einem der vorhergehenden Ansprüche, wobei die Messvorrichtung eine Vortexmessvorrichtung ist, und wobei das zeitabhängige Signal einen Abschnitt, der für einen Durchfluss repräsentativ ist, und einen Abschnitt, der für ein Rauschen repräsentativ ist, umfasst.
     
    6. Verfahren nach einem der Ansprüche 1 bis 4, wobei die Messvorrichtung ein Wärmedurchflussmesser ist, der eine Vielzahl von Temperatursensoren zur Messung eines Wärmeprofils um einen erhitzten Körper umfasst.
     
    7. Verfahren zur Bestimmung eines Durchflusses einer Vortexmessvorrichtung, umfassend den Schritt des Berechnens einer Sequenz von Frequenzspektren, wobei jedes Frequenzspektrum der Sequenz durch das Verfahren nach Anspruch 4 berechnet wird, wobei für das erste Spektrum der Sequenz die vorbestimmte Anzahl 1 ist, und für jedes nachfolgende Spektrum der Sequenz die vorbestimmte Anzahl um einen vorbestimmten positiven ganzzahligen Wert erhöht wird, und
    wobei für jedes Frequenzspektrum der Sequenz von Frequenzspektren der entsprechende Durchfluss durch Anwenden des vorbestimmten Algorithmus bestimmt wird, und
    des Evaluierens, ob der mindestens eine Durchfluss als größer als das Grundrauschen bestimmt wird, wobei dann das erste Spektrum der Sequenz von Spektren, bei denen der vorbestimmte Durchfluss größer ist als das Grundrauschen, als Basis zur Berechnung des Durchflusses der Vortexmessvorrichtung ausgewählt wird, ansonsten Bestimmen des Durchflusses als Null-Durchfluss.
     
    8. Verfahren nach Anspruch 7, wobei jedes Frequenzspektrum in eine Anzahl von diskreten Frequenzkomponenten geteilt wird, und das Identifizieren der Spitzenwertkomponente den Schritt des Ableitens der Spitze durch Interpolation auf der Basis von einigen Frequenzkomponenten umfasst.
     
    9. Verfahren nach Anspruch 7, wobei jedes Frequenzspektrum der Sequenz von Frequenzspektren mit einer konsekutiv aufsteigenden Kanalnummer nummeriert wird, wobei das erste Frequenzspektrum mit der kleinsten Kanalnummer nummeriert wird, das letzte Frequenzspektrum mit der größten Kanalnummer nummeriert wird, wobei Frequenzspektren, die höhere Kanalnummern aufweisen, über Frequenzspektren vorbestimmter Anzahlen von zeitabhängigen Signalen mit niedrigeren Frequenzgrenzen gemittelt werden als Frequenzspektren vorbestimmter Anzahlen von zeitabhängigen Signalen mit niedrigeren Kanalnummern,
    wobei vor dem Evaluierungsschritt von Anspruch 7 für jedes Frequenzspektrum der Sequenz von Frequenzspektren der entsprechende Durchfluss durch Anwenden des vorbestimmten Algorithmus bestimmt wird, und
    wenn der Durchfluss des Spektrums, das die kleinste Kanalnummer aufweist, als größer als das Grundrauschen bestimmt wird, so wird dieses Spektrum als Basis zur Berechnung eines Durchflusses der Vortexmessvorrichtung im stationären Zustand ausgewählt,
    wenn ansonsten der Durchfluss des Spektrums, das die kleinste Kanalnummer aufweist, als gleich dem Grundrauschen und mindestens einem der Spektren, die eine Kanalnummer zwischen der kleinsten und der größten Kanalnummer mit einem Durchfluss größer als das Grundrauschen aufweisen, bestimmt wird, eines der Spektren mit einem Durchfluss größer als das Grundrauschen als Basis zur Berechnung eines dynamischen Durchflusses der Vortexmessvorrichtung ausgewählt wird,
    ansonsten der Evaluierungsschritt von Anspruch 7 vorgenommen wird.
     
    10. Verfahren nach einem der vorhergehenden Ansprüche, ferner umfassend den Schritt der Bestimmung eines stationären Durchflusses und eines dynamischen Durchflusses.
     
    11. Verfahren nach einem der Ansprüche 1 bis 5 oder 7 bis 10, wobei der Ausgang der Messvorrichtung einen relativen Druck zwischen den mindestens zwei Sensoren repräsentiert.
     
    12. Verfahren nach einem der Ansprüche 5 oder 7 bis 9, wobei die Vortexmessvorrichtung einen Vortexsensor umfasst, der innerhalb eines Gehäuses mit einer Öffnung angeordnet ist.
     
    13. Vortexsensor, welcher dafür geeignet ist, das Verfahren nach einem der Ansprüche 1 bis 5 oder 7 bis 12 durchzuführen.
     


    Revendications

    1. Procédé pour évaluer un spectre de fréquences représentatif d'au moins un signal dépendant du temps, le signal dépendant du temps au nombre d'au moins un étant dérivé d'une sortie d'un dispositif de mesure sous des conditions de fonctionnement de dispositif de mesure prédéterminées, le signal dépendant du temps comprenant une partie qui est représentative d'un signal voulu et une partie qui est représentative du bruit, le procédé comprenant les étapes consistant à:

    - déterminer, en fonction du spectre de fréquences du signal, une valeur représentative du plancher de bruit,

    - identifier, en fonction du spectre de fréquences du signal dérivé sous la condition de fonctionnement prédéterminée, une composante de pic, et

    - si la composante de pic satisfait à un critère de pic relatif déterminé en fonction de la valeur déterminée représentative du plancher de bruit, déterminer le signal voulu par l'application d'un algorithme prédéterminé, dans lequel la valeur représentative du plancher de bruit est calculée sous la forme d'une moyenne d'un nombre de composantes de spectre de fréquences représentatives du bruit sélectionnées,

    caractérisé en ce que:
    les composantes de spectre de fréquences représentatives du bruit sont sélectionnées parmi des composantes de fréquence inférieures à une valeur de fréquence prédéfinie, les composantes de spectre de fréquences représentatives du bruit étant les composantes restantes après la sélection d'un nombre prédéterminé de composantes de fréquence de valeur élevée, les composantes de fréquence de valeur élevée étant un nombre prédéterminé de composantes de fréquence avec les valeurs d'amplitude les plus élevées.
     
    2. Procédé selon la revendication 1, dans lequel un pic de signal relatif est déterminé à partir de la différence entre le plancher de bruit et le signal voulu, le pic de signal relatif étant supérieur ou égal au critère de pic relatif.
     
    3. Procédé selon l'une quelconque des revendications précédentes, dans lequel l'algorithme prédéterminé est adaptatif vis-à-vis du bruit par le fait de:

    - toujours calculer le critère de pic relatif en fonction du bruit qui est présent, et

    - sélectionner une procédure de réalisation de moyenne avec un nombre de réalisations de moyenne, le nombre de réalisations de moyenne étant fonction du plancher de bruit actuel.


     
    4. Procédé selon l'une quelconque des revendications 1 à 3, dans lequel le spectre de fréquences est calculé par la réalisation d'une moyenne de spectres de fréquences d'un nombre prédéterminé de signaux dépendant du temps dérivés à partir de la sortie du dispositif de mesure.
     
    5. Procédé selon l'une quelconque des revendications précédentes, dans lequel le dispositif de mesure est un dispositif de mesure de tourbillonnement, et dans lequel le signal dépendant du temps comprend une partie qui est représentative d'un écoulement, et une partie qui est représentative du bruit.
     
    6. Procédé selon l'une quelconque des revendications 1 à 4, dans lequel le dispositif de mesure est un débitmètre thermique comprenant une pluralité de capteurs de température pour mesurer un profil thermique autour d'un corps chauffé.
     
    7. Procédé pour déterminer un écoulement d'un dispositif de mesure de tourbillonnement, comprenant l'étape de calcul d'une séquence de spectres de fréquences, chaque spectre de fréquences de la séquence étant calculé par le procédé selon la revendication 4, dans lequel, pour le premier spectre de la séquence, le nombre prédéterminé est 1, et, pour chaque spectre suivant de la séquence, le nombre prédéterminé est accru d'une valeur entière positive prédéterminée, et
    dans lequel, pour chaque spectre de fréquences de la séquence de spectres de fréquences, l'écoulement correspondant est déterminé par l'application de l'algorithme prédéterminé, et
    s'il est évalué qu'au moins un écoulement est déterminé comme étant supérieur au plancher de bruit, alors, le premier spectre de la séquence de spectres où l'écoulement déterminé est supérieur au plancher de bruit est sélectionné comme base pour le calcul de l'écoulement du dispositif de mesure de tourbillonnement, et, sinon, l'écoulement est déterminé comme étant un écoulement nul.
     
    8. Procédé selon la revendication 7, dans lequel chaque spectre de fréquences est divisé en un certain nombre de composantes de fréquence individuelles, et l'identification de la composante de pic comprend l'étape de dérivation du pic par interpolation en fonction de plusieurs composantes de fréquence.
     
    9. Procédé selon la revendication 7, dans lequel chaque spectre de fréquences de la séquence de spectres de fréquences est numéroté avec un numéro de canal croissant de façon consécutive, le premier spectre de fréquences étant numéroté avec le plus petit numéro de canal, le dernier spectre de fréquences étant numéroté avec le plus grand numéro de canal, des spectres de fréquences ayant des numéros de canal plus élevés ayant leur moyenne qui est réalisée sur des spectres de fréquences de nombres prédéterminés de signaux dépendant du temps avec des limites de fréquence inférieures à celles de spectres de fréquences de nombres prédéterminés de signaux dépendant du temps ayant des numéros de canal moins élevés, dans lequel, avant l'étape d'évaluation de la revendication 7, pour chaque spectre de fréquences de la séquence de spectres de fréquences, l'écoulement correspondant est déterminé par l'application de l'algorithme prédéterminé, et
    si l'écoulement du spectre ayant le plus petit numéro de canal est déterminé comme étant supérieur au plancher de bruit, ce spectre est sélectionné comme base pour calculer un écoulement à l'état stable du dispositif de mesure de tourbillonnement,
    sinon, si l'écoulement du spectre ayant le plus petit numéro de canal est déterminé comme étant égal au plancher de bruit et qu'au moins l'un des spectres ayant un numéro de canal entre le plus petit et le plus grand numéro de canal a un écoulement supérieur au plancher de bruit, l'un des spectres ayant un écoulement supérieur au plancher de bruit est sélectionné comme base pour calculer un écoulement dynamique du dispositif de mesure de tourbillonnement,
    et, sinon, il est effectué l'étape d'évaluation de la revendication 7.
     
    10. Procédé selon l'une quelconque des revendications précédentes, comprenant de plus une étape de détermination d'un écoulement stationnaire et d'un écoulement dynamique.
     
    11. Procédé selon l'une quelconque des revendications 1 à 5 ou 7 à 10, dans lequel la sortie du dispositif de mesure représente une pression relative entre les capteurs au nombre d'au moins deux.
     
    12. Procédé selon l'une quelconque des revendications 5 ou 7 à 9, dans lequel le dispositif de mesure de tourbillonnement comprend un capteur de tourbillonnement qui est disposé à l'intérieur d'un boîtier à un orifice.
     
    13. Capteur de tourbillonnement adapté de façon à mettre en oeuvre le procédé selon l'une quelconque des revendications 1 à 5 ou 7 à 12.
     




    Drawing














    Cited references

    REFERENCES CITED IN THE DESCRIPTION



    This list of references cited by the applicant is for the reader's convenience only. It does not form part of the European patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be excluded and the EPO disclaims all liability in this regard.

    Patent documents cited in the description